Favorite Audio Amp Design

This is simple, reliable and works very well:

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

(Might be out of your price range) The guy who makes these is less than a mile from where I work

brent

I like that circuit, but unfortunately it doesn't look like it could do

Is the diode backwards and should there be two of them?

John

Ah, yes! What we did with the MC1554G back in the early '60's ;-) ...Jim Thompson

I don't think the diode is there for biasing purposes, I think it's used to provide drive current for the output transistors through the first stage. Otherwise the load resistors on the first stage would have to be sized to be able to provide the peak output base current for the output stage without appreciable voltage drop, which would increase quiescent dissipation.

"Vladimir Vassilevsky"

At idle, the diode is reverse biased by 1.3 volts and cannot conduct drive current into the output devices until the op-amp swings +/-1.3 volts *more* than at the output point. This will happen at some output current level depending on the un-specified resistors' ability to supply drive current. The take over will only be smooth if a lot of unspecified things are got right.

Also, the arrangement is unable to drive the output devices fully whenever the output current and voltage are not the same polarity - ie with a reactive load like a speaker.

Idle bias for the output pair depends on the relative Vbes of the drive and output devices, it is not compensated for temp variations in the output pair and is not stabilised by emitter ballast resistors. Bias will increase dramatically as the output pair heat and may go into thermal runaway.

Because of all the Vbes, the diode and op-amp losses at each rail - the available output swing into a 4 ohm load is about 3 to 4 volts less in each polarity than the DC rails, which is absurd for a stage that must run from an overall 12 volt supply.

.... Phil

I don't recall the exact circuit, but I can tell you my least favorite audio circuit was the one Motorola used in their Pagecom tone-&-voice VHF beepers. It was a discrete 7-transistor DC-coupled monstrosity, which was an absolute bitch to troubleshoot if one of the transistors died. Of course, it was pretty loud and ran off two 1.5 volt N-size batteries.... In 1980.

-mpm

Nice, I don't even have to copy that down. Thanks,

George H.

How about a minimal solution, just an opamp to give the V gain, followed by an unbiased npn/pnp output pair, with the nfb inlcluding the output devices in the loop. If the parts are reasonably fast (far above af) and the output devices smallish, it can correct for crossover distortion quickly, minimising total distortion. (With a slow opamp it sounds horrid.) Minimal parts count, trivially easy to troubleshoot and replace with other parts later, minimum cost, good energy efficiency for a linear amp.

NT

Here's a circuit, that while it isn't particularly innovative, will easily do over 3 watts into a 4 ohm load from a 12 volt supply:

The two big blobs are each an LM386, which has been around forever and will likely be in production for some time. The LM386 has enough drive capability to power the output transistors even if they should have low beta, and also biases up to half the supply voltage by itself. With its internal feedback resistor bypassed it also has enough gain to reduce the crossover distortion significantly. Two of these stages bridged can do about 4.5 watts into 4 ohms at clipping.

The distortion at 10 kHz doesn't look too horrible:

If you don't mind your LM386es in DIP packages and get them here:

for 18 cents each, the price of the whole amplifier would probably be less than a dollar fifty.

Did I just miss something or does this require a split supply? I've only got unipolar +12 to work with.

Jim

As I recall the 1554 would sing like a bird on the slightest provocation and go phffft if spoken to sharply.

{;-)

Jim

I'm trying to visualize what you have written and can't come up with a reasonable design. Care to post a drawing somewhere?

Thanks,

Jim

But some day they WILL be out of production and go into unobtanium land. Right now I've got several tens of thousands unrepairable units in the field with "around forever" chips in them like the Fairchild uA706 that are absolutely gone, the TBA 801 that can be had in the aftermarket for $$$, and so on.

Nope, jelly bean opamps and discretes, both of which will be around when the 386 is in the dumper.

Thanks,

Jim

I'm guessing similar to each half of what bitrex posted:

-Lasse

If you tie the bases and the emitters together as you suggest, it's simple and thermally stable but has crossover distortion. One simple fix is to add one b-e resistor, 100 ohms maybe. At low signal swings, the opamp drives the load directly through the resistor. When the drop across the resistor hits 0.7 or so, the transistors pick up the load. The open-loop transfer function is nonlinear but continuous, so a bunch of NFB can hammer it flat.

Another cute trick is to have the opamp drive the load directly, but add PNP and NPN "helper" transistors with emitters on the power rails and whose bases are driven by the power pins of the opamp. There's a mosfet variation of that, too.

John

We sometimes make small pcb adapters as drop-ins for chips we can't get any more. We had to make several thousand of these when Maxim discontinued the MAX9690 without notice.

ftp://jjlarkin.lmi.net/Break2.jpg

ftp://jjlarkin.lmi.net/OnBoard.jpg

John

Yes, similar to that.

(For one offs you can even use a slow opamp and germaniums to avoid crossover distortion.)

NT